PowerMac G5 context: kernel_va_to_slbv(...) and SLBV_L use vs. addresses < DMAP_BASE_ADDRESS

[DMAP_START below is an abbreviation of DMAP_BASE_ADDRESS.]

kernel_va_to_slbv has logic for if slbv would have
SLBV_L asserted or not:

uint64_t
kernel_va_to_slbv(vm_offset_t va)
{
        uint64_t slbv;

        /* Set kernel VSID to deterministic value */
        slbv = (KERNEL_VSID((uintptr_t)va >> ADDR_SR_SHFT)) << SLBV_VSID_SHIFT;

        /*
         * Figure out if this is a large-page mapping.
         */
        if (hw_direct_map && va > DMAP_BASE_ADDRESS && va < DMAP_MAX_ADDRESS) {
                /*
                 * XXX: If we have set up a direct map, assumes
                 * all physical memory is mapped with large pages.
                 */
        
                if (mem_valid(DMAP_TO_PHYS(va), 0) == 0)
                        slbv |= SLBV_L;
        }
        
        return (slbv);
}

[Presuming hw_direct_map!=0 . . .]

By having handle_kernel_slb_spill count its usage,
broken out by address range and EXC_[DI]SE type, there
are examples of slbtrap's that use handle_kernel_slb_spill
for srr0<DMAP_START and for dar<DMAP_START.

handle_kernel_slb_spill uses kernel_va_to_slbv. (So did
my hack to -r345758 .)

Are slbv values correct for small effective addresses by
not having SLBV_L asserted? (Same memory, covered by two
different page sizes based on the style of
effective-address that happens to be used [upper bits
different].)



===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)

_______________________________________________
[email protected] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-ppc
To unsubscribe, send any mail to "[email protected]"